MIT OpenCourseWare
http://ocw.mit.edu
HST.583 Functional Magnetic Resonance Imaging: Data Acquisition and Analysis
Fall 2006
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
HST.583: Functional Magnetic Resonance Imaging: Data Acquisition and Analysis, Fall 2006
Harvard-MIT Division of Health Sciences and Technology
Course Director: Dr. Randy Gollub.
Laboratory 2: Introduction to fMRI Data and Analysis
September 18th, 2006
Divya Bolar
Exercises and Questions
1. After overlaying the activation map on the anatomical/structural volume in exercise 4
(page 25), adjust the threshold and opacity parameters such that only activated regions of
the overlay can be seen. What parameter values did you use?
2. Do the activations appear to be in the right places (roughly) based on how the visual
fields map to the visual cortex? Why or why not?
3. Instead of using [1 -1] to set the Left - Right contrast, use [1 0] and then [0 1] when
applying the linear model. What do these activation maps tell you? How is this
information different from the map produced with [1 -1]? What if [-1 1] is used? How is
the -log(p) map different in this case? Does this contrast setting provide any additional
useful information?
4. Would it be correct to just use one of the two design blocks to try to fit the data? What
would such a map tell you? In other words, if just the left hemifield design block was
used for linear modeling, what would be the significance of the resulting -log(p) map?
5. Give two reasons why it is beneficial to spatially smooth before performing a general
linear model analysis? Would it still be useful if fMRI activations were not spatially
correlated? Why or why not?
6. What is the major drawback of spatially smoothing?
7. Perform GLM analysis without spatially smoothing. Are the results consistent with what
you would expect?
8. Perform GLM analysis without performing any motion correction. What are the major
differences in the activation maps?
9. On the "Model" page of the Linear Modeling action, it is possible to remove polynomial
drift, up to the third order. What is the benefit of having a non-zero drift order? Extra
credit: Give an example when using a non-zero drift order may not benefit the analysis.
10. Change the original EPI time series (i.e. ge_functionals8_64) to a spectral colormap.
Then, overlay this series (not the -log(p) map) onto the structural. Make the opacity of
the overlay about 0.35 and take a look at a mid-cortical axial slice. Do you notice any
spatial differences between the EPI image and the anatomical image? Extra credit: Why
is this occurring and how could it be minimized? (hint: check the types of scans acquired
in lab 1; review the purpose of each one).
Cite as: Divya S. Bolar, HST.583 Functional Magnetic Resonance Imaging: Data Acquisition and Analysis, Fall 2006.
(Massachusetts Institute of Technology: MIT OpenCourseWare), http://ocw.mit.edu (Accessed MM DD, YYYY). License: Creative
Commons BY-NC-SA.